This is sort of a follow up question to this question on Satake-Tits diagrams.
In the question, user Callum comments that for, $\mathfrak{so}(n, \mathbb{C})$, the real forms (potentially with double counting up to algebra homomorphism) are given by $\mathfrak{so}(p,q, \mathbb{R})$ with $p+q=n$, and then if $n$ is even also $\mathfrak{so}^*(n)$. In particular I'm assuming we are considering real Lie algebra's over $M_{n\times n}$ for entries in either $\mathbb{R}$, $\mathbb{C}$, or $\mathbb{H}$.
My question is twofold:
1. It does not seem obvious why there couldn't be $\mathfrak{so}^*(p,q)$. Do these not form a group, or do they always coincide with another real form? ANSWER: Coming back to this now, it is clear Sylvester's law of inertia does not hold for these preserved structures, so $\mathfrak{so}^*(p,q)$ is always a change of basis away from deserving the name $\mathfrak{so}^*(n)$.
2. How do we prove that this is all of the matrix groups which, after complexification, give the relevant group? (This question is not necessarily about the orthogonal groups in particular but real forms of any and all semi-simple Lie algebra's)
In small enough dimension it makes sense to do something like an exhaustive search. However given my surprise at the existence of the quaternionic orthogonal algebra it seems likely counting signature's alone as my intuition tells me, doesn't guarantee I'll find them all, especially as the algebra's get larger.
My current approach is that if there is a signature to play with, each one will be a real form. Then, look to covering maps (in either direction) and see if playing with the signature there gives anything new. Lastly, special cases of quaternionic real forms, and 'scalar restriction' of a complex group. But this is not as systematic as I would like, and leaves me unconvinced I have 'every real form'.